---

name: qa description: "QA pipeline: explores the running app like a real user using a browser, then produces a QA report and Playwright E2E tests for regression. Orchestrates app-scout → qa-agent." argument-hint: "[--fresh] [scope or feature to test — leave empty to test everything]"

QA Pipeline

You are orchestrating the QA pipeline. Follow these steps strictly in order.

Input

The test scope (optional — empty means test everything):

$ARGUMENTS

Step 0: Resolve QA_OUTPUT_DIR + clean up

Step 0a: Resolve QA_OUTPUT_DIR

Run the following in Bash to determine the branch-scoped QA output directory:

RAW_BRANCH=$(git rev-parse --abbrev-ref HEAD 2>/dev/null || echo "")
if [ -z "$RAW_BRANCH" ] || [ "$RAW_BRANCH" = "HEAD" ]; then
  SHORT_SHA=$(git rev-parse --short HEAD 2>/dev/null || echo "")
  RAW_BRANCH=${SHORT_SHA:+detached-$SHORT_SHA}
fi
if [ -z "$RAW_BRANCH" ]; then
  echo "Warning: not a git repo — using qa-output/ as QA output directory" >&2
  QA_OUTPUT_DIR="qa-output"
else
  SANITIZED=$(echo "$RAW_BRANCH" | tr '/' '-' | tr -cs 'A-Za-z0-9._-' '-' | sed 's/^-*//; s/-*$//')
  [ -z "$SANITIZED" ] && SANITIZED="unknown-branch"
  QA_OUTPUT_DIR="qa-output/$SANITIZED"
fi

Store QA_OUTPUT_DIR as a session variable — use it everywhere below.

Step 0b: Clean up stale QA artifacts

Use Bash to run rm -rf "$QA_OUTPUT_DIR" to clear only this branch's QA artifacts from a previous run.

Step 0.5: App Recon — Discover how to interact with the app

Parse flags: If $ARGUMENTS starts with --fresh, set FRESH=true and strip --fresh to get the clean scope. Otherwise FRESH=false. Store the cleaned scope as QA_SCOPE; use QA_SCOPE everywhere below instead of raw $ARGUMENTS.

Check whether to run app-scout:

• Run via Bash: find .claude/app-context.md -mmin -60 2>/dev/null
• If the file path is returned (exists and < 1 hour old) AND FRESH=false: Use it directly — skip launching app-scout. Proceed to Step 1.
• Otherwise (file missing, stale, or --fresh was passed): Launch the app-scout agent using the Task tool with:
  • subagent_type: "app-scout"
  • Prompt: Perform project recon. Write your findings to .claude/app-context.md.

Wait for it to complete. If the agent fails or the file is not created, log a warning and proceed without it — this is a best-effort step.

Step 1: QA — Test the app and generate outputs

Read .claude/app-context.md (from Step 0.5). Build the qa-agent prompt as follows (if .claude/app-context.md does not exist, omit the App Context section but always include QA_OUTPUT_DIR):

Test scope: $QA_SCOPE (if empty: test all major flows)

QA_OUTPUT_DIR=$QA_OUTPUT_DIR

Write all outputs (report, screenshots, etc.) under this branch-scoped directory.

## App Context (from pre-recon)

The following was pre-discovered about this project. Use the start command and
URL from this context. Re-check running status yourself.

<full content of .claude/app-context.md>

Launch the qa-agent using the Task tool with:

• subagent_type: "qa-agent"
• Provide the constructed prompt above (with $QA_OUTPUT_DIR expanded)

Wait for it to complete.

Step 2: Report

Read $QA_OUTPUT_DIR/qa-report.md and summarize to the user:

## QA Complete

**Scope:** <what was tested>

**Results:** N passed · N failed · N warnings

**Critical issues:**
- [list from report, or "None found"]

**Outputs:**
- Full report: $QA_OUTPUT_DIR/qa-report.md
- E2E tests: <list files written>
- Run tests: playwright test

If $QA_OUTPUT_DIR/qa-report.md does not exist, report that the QA agent failed to complete and show any error output.

Rules

• Run steps sequentially — each depends on the previous
• If Step 0.5 fails (no app-context.md created), log a warning and continue to Step 1 anyway — qa-agent will discover the app independently; this is non-fatal
• If the app is not running, the qa-agent will report it — do not attempt to start the app here
• Always present the summary in Step 2, even if QA found no failures